AI agents: From weekend project to global race

The AI agent landscape is experiencing a significant resurgence, with platforms like Meta's Muse and Instinct gaining traction. Muse, recently topping app store charts, reportedly has 600,000 daily active users in the US, while Instinct is valued at $2.5 billion. Both platforms exhibit similarities to OpenClaw, the project widely credited with popularizing AI agents.

Concerns have been raised that Meta's Muse may be built upon OpenClaw, citing identical core file names and similar lines of code governing AI personality. Meta has denied direct copying, stating Muse was built 'from scratch' but admitted it was 'heavily inspired' by OpenClaw, with Meta's product head purchasing hundreds of Mac Minis to facilitate its development.

Instinct, while less directly derivative, offers messaging-based interaction similar to OpenClaw's revolutionary features, enabling users to delegate personal assistant tasks. These use cases, ranging from appointment booking to recovering lost items, mirror OpenClaw's initial appeal.

OpenClaw, a one-man weekend project, transformed AI agents from concepts to practical tools within ten months. Its success, attracting millions of visitors and thousands of GitHub stars, spurred significant investment and attention from major AI companies. OpenAI hired its creator in February, followed by Google's and Apple's major announcements regarding agentic AI in May and June, respectively.

Instinct's rapid growth and substantial funding round in August, preceding Meta's Muse introduction in September, highlight the intense competition. The core question now is whether these new platforms have improved upon OpenClaw, particularly addressing its significant security flaws, which included malware in downloaded skills.

Meta emphasizes Muse's privacy and security features, storing data in an isolated VM. However, Meta retains access to user data, with plans to implement stronger protections later this year. Muse also defaults to using user data for model training, though opting out is possible. A recent zero-day vulnerability has also been flagged, allowing agent hijacking.

Instinct has faced criticism regarding its terms of service, though adjustments appear to have been made. Ultimately, both Meta and Instinct are making AI agents more accessible. Muse integrates with Meta's ecosystem, while Instinct's messaging interface broadens its appeal. Their lower cost compared to competitors and simplified download process for Muse further reduce barriers to entry.